How Much Power Does a Jetson Orin Nano Use Per Camera?

For complete Orin Nano power consumption benchmarks across all modes (idle, 15W, MAXN Super), see Jetson Orin Nano Power Consumption: Idle, 15W, MAXN Benchmarks.

Power Per Camera Table

The table below shows estimated total system power for the Jetson Orin Nano at different camera counts, resolutions, and model sizes. All measurements use TensorRT INT8 inference at 15 fps per stream unless noted otherwise.

Planning estimates — use the power calculator for exact values. All figures represent module power only; add peripheral power (cameras, NVMe, radios) for full system budget.

How to read this table

The first four rows show the scaling curve for the most common deployment scenario: 1080p cameras with a lightweight detection model. Notice that power does not scale linearly with camera count. The first camera triggers GPU clock ramp-up from idle (5W to 6–8W), but each subsequent camera only adds 1–2W because the GPU and memory subsystem are already active.

The bottom rows show what happens when you increase model complexity (YOLO11s, YOLO11m), resolution (4K), or camera count beyond the 15W mode's comfortable range. These scenarios require MAXN mode and active cooling.

What Affects Power Per Camera

Five variables determine how much power each camera adds to the system. Understanding these lets you predict whether your deployment fits within a given power envelope.

Resolution

Doubling resolution quadruples the pixel count and roughly doubles the memory bandwidth required for frame decode and tensor preprocessing. Moving from 1080p to 4K adds 3–5W to total system power for the same camera count and model. If your detection task does not require 4K detail — and most perimeter security and people-counting tasks do not — downscaling to 1080p before inference is the single most effective power optimization.

Model size

Model complexity directly correlates with GPU utilization and power draw. YOLO11n (nano) has 2.6M parameters; YOLO11s (small) has 9.4M; YOLO11m (medium) has 20.1M. Each step up roughly doubles the per-camera power cost. For multi-camera deployments on a power budget, start with the lightest model that meets your accuracy requirements.

Precision

INT8 quantization halves the memory bandwidth and compute cycles compared to FP16, and quarters them versus FP32. On the Orin Nano, INT8 inference at 1080p with YOLO11n draws 6–8W; the same workload at FP32 draws 9–12W. TensorRT handles the quantization automatically with a calibration dataset.

Frame rate

Power scales sub-linearly with FPS. Dropping from 30 fps to 15 fps per stream reduces GPU utilization by roughly 40–50% and saves 1–3W per camera depending on model size. For many surveillance and monitoring applications, 10–15 fps provides sufficient temporal resolution while significantly reducing the power envelope.

Power mode

The Orin Nano's nvpmodel power modes cap CPU and GPU clock frequencies. The 15W mode is the default for production deployments and handles 4–6 cameras at 1080p comfortably. MAXN Super removes the power governor, allowing the GPU to boost freely — necessary for 8+ cameras or heavier models, but requiring active cooling and a larger PSU.

→ Full 15W vs MAXN Super comparison with benchmarks →

PoE Power Budget Implications

If you are powering the Jetson Orin Nano over PoE (via a carrier board with PoE input), the available power at the device determines your maximum camera count.

Key constraint: 802.3af delivers only 12.95W at the powered device. After DC-DC conversion losses (~15%), you have roughly 11W available — enough for 2–3 cameras with YOLO11n. For 4+ camera deployments, you need 802.3at (PoE+) at minimum.

Remember that these figures cover only the Orin Nano module power. The cameras themselves draw power separately from the PoE switch — typically 7–15W per camera depending on whether they have IR illuminators, heaters, or PTZ motors. Budget the switch's total PoE budget accordingly.

→ Calculate your full PoE power budget →

Battery and Solar Deployments

For mobile, temporary, or off-grid deployments, battery capacity determines how long the Orin Nano can run at a given camera count.

Runtime estimates

Assumes module power only, no peripheral draw. Multiply by 0.85 for real-world efficiency losses (DC-DC conversion, battery degradation).

Extending battery runtime

• INT8 quantization — reduces average draw by 20–30% versus FP16
• Motion-triggered inference — run detection only when the camera detects motion, dropping average draw toward idle (5W) during quiet periods
• Lower FPS — 10 fps instead of 30 fps saves 1–3W per camera
• Fewer cameras during off-peak — disable streams programmatically to reduce GPU load

Solar sizing

For a 4-camera deployment averaging 11W, a 30W solar panel with a 50 Wh battery provides roughly 8–12 hours of autonomous operation per day in moderate sunlight (4–5 peak sun hours). Oversizing the panel by 2–3x accounts for cloudy days, panel degradation, and the mismatch between solar availability and inference demand.

Calculate Power for Your Exact Setup

The estimates above cover common configurations. For exact numbers tailored to your deployment — including peripherals, cooling, and PoE overhead — use the interactive tools:

Frequently Asked Questions

Related Guides

• Jetson Orin Nano Power Consumption: Idle, 15W, MAXN Benchmarks
• Jetson Orin Nano 15W vs MAXN Super: Which Power Mode to Use
• Jetson Orin Nano vs NX Power Consumption Compared
• Edge AI Module Power Calculator
• Edge AI System Designer